- Nathan Besh, Cost Lead, Well-Architected
- Spencer Marley, Commercial Architect
- Paul Lambden, Principal Technical Account Manager
If you wish to provide feedback on this lab, there is an error, or you want to make a suggestion, please email: [email protected]
- View an RI report
- Download and prepare the RI CSV files
- Sort and filter the RI CSV files
- Teardown
- Rate this Lab
We are going to view the RI reports within AWS Cost Explorer, to understand the recommendations and possible purchases we should make.
-
Log into the console as an IAM user with the required permissions, go to the AWS Cost Explorer service page:
-
In the left menu select Recommendations:
-
Click View all next to Reservation purchase recommendations:
-
On the right select the filters: RI term 1 year, Payment Option (your preference), Based on the past 7 days:
The top section will show the estimated savings and number of recommendations, take note of the Purchase Recommendations
- On the right select the filter: Based on the past 30 days:
View the Purcahse Recommendations, if the 30 day recommendation is less than 7 days recommendation - your usage is increasing and the recommendations are lower risk. If the 7 days recommendation is less than 30 days, then your usage is decreasing and you need to look further into your usage patterns to see which RI's would be suitable.
1 - Download the CSV for both the 7 day and 30 day recommendation files, by selecting the filter 7 days or 30 days, and clicking on Download CSV:
2 - If you do not have sufficient usage, you can download the two sample files:
Ctrl-click to open them in a new tab, then copy the text and paste it into a spreadsheet application. Paste the 30day recommendations into one worksheet, and the 7day recommendations into another worksheet called 7Day Rec, in the same spreadsheet.
3 - Create a new column called RI ID to the left of the Recommendation column on both 30Day and 7Day sheets, which is a unique identifier of the RI Type, the formula for this cell will concatenate the columns: Instance Type, Location,OS and Tenancy. On row 5 of the sample files, paste the formula below. NOTE If using your own files modify the row numbers from 5 to 2.
=CONCATENATE(C5,L5,M5,N5)
4 - Add a column in the 30Day worksheet to the right of the Recommendation column. This will be for the 7Day recommendations. Add a VLOOKUP formula to get the values from the 7Day worksheet, modify this formula for the number of rows you require (U column). Paste this on row 5 of the sample files, or modify if using your own:
=VLOOKUP(T5,'7Day Rec'!T$4:U$48,2,FALSE)
5 - We will now create a Fully Paid Day column. This shows us how long it will take to pay off the full term of the RI, and will help to measure risk. The closer to 12months the fully paid day is, the higher the risk. This is because it takes longer to pay off the investment/commitment. The break even is the wrong measure, as it only shows how quickly you pay off the upfront component, and not the full amount. Paste the following formula into the last column z:
=(R5+S5*12)/(R5/12+S5+W5)
The formula is:
(yearly RI cost) / (monthly on-demand cost)
(Upfront cost + recurring monthly cost x 12) / (upfront cost/12 + recurring monthly cost + estimated monthly savings)
6 - Delete the following columns as they are not necessary: Recommendation Date, Size Flexible Recommendation, Max hourly normalized unit usage in Historical Period, Min hourly normalized unit usage in Historical Period, Average hourly normalized unit usage in Historical Period, Payment Option, Break Even Months.
We now have the required data required to be able to analyze, and filter out the high risk and low return RIs.
RI purchases should be done frequently (bi-weekly or monthly), so for each cycle we want: low risk and high return purchases, and purchase the top 50-75% of recommendations. This will ensure you have sufficiently high coverage, while minimizing the risk of unused RIs.
1 - To get the lowest risk, we sort by Fully Paid Day smallest to largest, as these will be fully paid off in the shortest amount of time. You can see that some of the RI's below are fully paid off in around 7months - so if they are used for 7 months - they have paid themselves off completely.
2 - We will separate the very low, low, and medium risk recommendations. Add in some empty lines between Fully Paid Day of 8, 10, and copy the header line across:
3 - We have categorized the risk, so we will now look for the highest return recommendations in each category. Sort each of the three groups by Estimated Monthly Savings, largest to smallest:
4 - Depending on your usage and business, chose a minimum estimated monthly savings - a typical value for larger customers is in the range of $50-100. While they save money, these recommendations do not save enough - aim for the top 50-70% of recommendations. We have chosen $100, grey out anything less than this:
It would be a large amount of effort to view the daily usage patterns over the month for every recommendation - checking for declining usage or erratic usage, but we can do this programatically. By looking at the columns, we can assess the underlying usage pattern.
1 - If the Max hourly usage is close to Min hourly usage, within 75-100% - then the usage would be relatively flat, with low variance. Go through and highlight these cells green. You could do this with a formula, but a very fast judgment is ok:
2 - If the Average hourly usage is close to the Max hourly usage, then the minimum was only a small duration, so highlight anything green where the Average is roughly within 75-100% of the Max:
3 - Minimum utilization required for an RI varies by the discount level. The lowest discount level is approximately 20%, so we would look for a minimum utilization of >80%. While this is reflected through the Fully Paid Day (if utilization is low, Fully Paid would be very late), we'll double check & filter out only the very high utilization. Highlight anything above 90% in green:
4 - Now we look for a declining usage pattern. If the recommendation for the last 7 days is less than the 30 days, usage is declining - and you should consult your business to determine if usage will continue to fall. If the 7day Recommended Instance Quantity is equal or more than the 30day Recommended Instance Quantity then highlight the cell green:
5 - Now we will see if the recommendation is close to the average, if its not then usage is varying. If the recommendation is NOT above, equal or close to the average (within 10%) then remove the highlighting from the recommendation column:
The processed sample files are available here:
We look at each of the risk categories as follows:
1 - Very low and low risk
- For any recommendations that are highlighted in the 7Day column, recommend the lowest of the 30Day or 7Day Columns.
- For any recommendations that are highlighted in the Average hourly usage and Projected RI Utilization, select a percentage of either the 30Day or 7Day column (which ever is lower).
2 - Medium risk
- From the recommendations highlighted in the 7Day column, select a portion of these on a case by case basis based on business knowledge
Other suggestions for recommendations that do not fall into the categories above:
- Re-evaluate in another 7-14 days to observe the usage trend
- Purchase a lower percentage of the average hourly
- Purchase a higher percentage of the minimum hourly
You have successfully filtered and processed all the recommendations. You can now make low risk and high return recommendations that are suitable based on your ongoing usage patterns with high confidence. You can then take those recommendations, and purchase the quantity in the required accounts, with the required payment option (All upfront, Partial upfront, No upfront), and class (standard or convertible).
There are no resources or configuration items that are created during this workshop.
